What is a Large Language Model (LLM)?
A large language model (LLM) is an AI model trained on enormous amounts of text to predict which words come next, which lets it answer questions, summarize and write in natural language. LLMs are the engines behind AI assistants such as ChatGPT, Claude, Gemini and Perplexity.
During training, the model reads a vast mix of training data, such as web pages, books and code, and learns patterns: which words, facts and brands tend to appear together. It does not store pages like a database. It stores statistical associations, which is why it can write fluently about your category and still get a detail about your company wrong.
Many assistants now combine the model with live web search. With retrieval-augmented generation, the assistant first fetches current pages and then lets the LLM write an answer based on them, often with links to its sources.
Why LLMs matter for your brand
When a buyer asks an AI assistant for "the best invoicing tool for freelancers", an LLM decides which brands to name, in what order and in what tone. That answer often takes the place of a list of search results, so the model's view of your brand becomes part of your reputation. If you are missing from its answers, you are missing from the shortlist.
Each LLM is trained on different data and draws on different sources, so ChatGPT, Gemini and Claude can describe the same company in different ways. Checking one assistant tells you little about the others. For a closer look at the selection process, read how AI models decide which brands to recommend.
How to influence what LLMs say about you
You cannot edit a model directly, but you can shape the information it learns from and retrieves:
Publish clear, factual pages about what you offer, who it is for and what it costs, in plain sentences a model can quote.
Earn mentions on independent sources that models trust, such as reviews, comparison articles and trade media.
Make sure AI crawlers can read your pages without JavaScript, so assistants that search the web find current information.
Keep your company facts consistent everywhere, which lowers the risk of an AI hallucination.
Test the prompts your buyers use in several assistants and repeat the test regularly, since answers change.
Trackbase runs your prompts every day across 11 AI models, including ChatGPT, Claude, Gemini, LLaMA and Mistral, so you can see where each one mentions you. The full list is on the integrations page.
Examples of large language models
GPT models (OpenAI) power ChatGPT. When a B2B payroll software vendor asks ChatGPT for "payroll software for small businesses", a GPT model decides which vendors make the shortlist.
Gemini (Google) powers the Gemini app and Google AI Mode, so Gemini's view of a Dutch bicycle webshop can shape both chat answers and search results.
Claude (Anthropic) runs in the Claude app and in many business tools, where it might summarize the pros and cons of competing accounting platforms.
Open models such as Meta's LLaMA and some Mistral models can be downloaded and built into other companies' chatbots and search tools, which spreads the same knowledge across many products.
Frequently asked questions
Is ChatGPT a large language model?
Not exactly. ChatGPT is an AI assistant, and the large language model is the engine inside it. ChatGPT runs on OpenAI's GPT models and adds features around them, such as web search, memory and file uploads. Other assistants work the same way: the app is the product, and the LLM is the model that writes the answers.
How do LLMs decide which brands to mention?
An LLM names the brands it most strongly associates with the question, based on patterns in its training data and, when it searches the web, on the pages it retrieves. Brands that are described often, consistently and positively on trusted sources are more likely to be mentioned. The position and wording can differ between models and from day to day.
Do all AI assistants use the same LLM?
No. ChatGPT uses OpenAI's GPT models. Gemini and Google AI Mode use Google's Gemini models, and Claude uses Anthropic's models. Some products, such as Perplexity, let users choose between several models. Because each model has different training data and sources, the same question can produce different brand recommendations, so check your visibility per assistant.